Review advertising recommendations with business guardrails

Treat a recommendation as a proposed change

An account suggestion can identify a useful opportunity, but it cannot know every operational limit of a small business. Review what the recommendation would change and why it matters to your current goal. Separate the platform's opportunity from your available appointment capacity, service area, and cash exposure. The useful question is whether the change supports suitable work under conditions the business can handle, not whether accepting it makes a score look more complete.

Source checks: Manage auto-apply recommendations · About change history

Inspect automatic subscriptions independently

Google provides settings for recommendation types subscribed to auto-apply, along with history and opt-in information. Review those settings even if the visible recommendation cards look reasonable. A subscription can affect future changes beyond today's card. Write who can enable or disable each type and which changes require business review. Do not assume that the absence of a recent manual edit means the campaign cannot have changed.

Use a one-page decision record

For each material proposal, record the campaign, current setting, proposed setting, intended outcome, possible mismatch, and review owner. Add a spend-exposure note where the change could affect traffic or cost. Include a rollback or revision plan when appropriate. Keep the record short enough to use during regular account reviews. The same process works for a manager's suggestion and an automated recommendation, because the business consequence matters more than its source.

Check service scope and signal quality first

New keyword or targeting suggestions need to fit what the team genuinely provides. Goal or bidding suggestions need reliable conversion definitions and values. Creative suggestions need accurate service promises and destinations. If the account counts repeated contact clicks as successful leads, a targeting expansion may optimize toward the wrong behavior. Fix the input problem before interpreting a recommendation's projected benefit as a business forecast.

Avoid bundling unrelated changes into one verdict

Applying several large suggestions together makes it difficult to tell which influenced the result. Where practical, change one material variable and preserve the prior configuration. An available experiment can support a controlled comparison, but a low-volume account may still yield inconclusive evidence. Immediate corrections to inaccurate claims or broken destinations need not wait for a growth test; identify those separately as repairs rather than marketing experiments.

Review what was actually applied

At the next check, inspect change history and auto-apply history alongside qualified inquiries, completed jobs, and spend. Confirm that the implemented change matches the decision record. Record unexplained differences and assign a person to resolve them. A decline or deferral is a valid outcome when the proposal does not fit the business. The goal is accountable account maintenance, with a clear link from each accepted change to the problem it was intended to solve.

Worked example: illustrative figures

For example, the account suggests adding three service keywords. The owner confirms two are offered and one requires a service the team does not provide. The two valid additions are reviewed against the destination and conversion goal; the third is declined. The account also has an auto-apply subscription the owner had not documented, so its scope is reviewed separately. Three suggestions with two accepted means two proposed inputs passed this business review, not a measured 66.7% performance gain or proof that those keywords will produce customers.

Questions about this task